Unveiling Urban Parking: Enhanced Image Analysis for Vacant Space Detection

Forked from official repository for the Image-Based Parking Space Occupancy Classification: Dataset and Baseline paper.

We introduce an effective training method using a custom data augmentation strategy to improve previous models' abilities to reason about occlusions caused surrounding vehicles, trees, and other natural or urban features.

In this repository, we provide:

• Code to reproduce all results.
• Download link for the dataset.
• Notebooks to explore the dataset, trial and visualize a variety of augmentations, train model, and visualize/ reproduce results.
• Scripts to alter elements of the model training process.

Dataset

The dataset (Action-Camera Parking Dataset) contains 293 images captured at a roughly 10-meter height using a GoPro Hero 6 camera. Here is a sample from the dataset:

Results

To reproduce our quantitative results, simply clone the repo and run the results notebook locally.

Training

To understand the workflow of the repo, train your own models, etc., we recommened working through the notebooks in the following order:

• acpds_playground
• roi_analysis
• augmentation_playground
• train
• results

Citation

@misc{marek2021imagebased,
      title={Image-Based Parking Space Occupancy Classification: Dataset and Baseline}, 
      author={Martin Marek},
      year={2021},
      eprint={2107.12207},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}